Unrestricted force

[michaelpj]

Don't look here yet. This is an experiment I've wanted to do for a
while, just looking to get some benchmark numbers.

Pre-submit checklist:

• Branch
  • Tests are provided (if possible)
  • Commit sequence broadly makes sense
  • Key commits have useful messages
  • Changelog fragments have been written (if appropriate)
  • Relevant tickets are mentioned in commit messages
  • Formatting, PNG optimization, etc. are updated
• PR
  • (For external contributions) Corresponding issue exists and is linked in the description
  • Targeting master unless this is a cherry-pick backport
  • Self-reviewed the diff
  • Useful pull request description
  • Reviewer requested

[michaelpj]

/benchmark plutus-benchmark:nofib

[github-actions]

Comparing benchmark results of ' plutus-benchmark:nofib' on '9fabd96d2' (base) and '6414603b4' (PR)

Results table

Script	9fabd96	6414603	Change
clausify/formula1	19.24 ms	19.22 ms	-0.1%
clausify/formula2	24.53 ms	24.38 ms	-0.6%
clausify/formula3	67.09 ms	66.40 ms	-1.0%
clausify/formula4	95.43 ms	94.18 ms	-1.3%
clausify/formula5	424.7 ms	419.7 ms	-1.2%
knights/4x4	54.70 ms	53.41 ms	-2.4%
knights/6x6	143.7 ms	142.0 ms	-1.2%
knights/8x8	236.8 ms	234.7 ms	-0.9%
primetest/05digits	27.67 ms	27.46 ms	-0.8%
primetest/08digits	49.45 ms	49.46 ms	+0.0%
primetest/10digits	69.15 ms	68.86 ms	-0.4%
primetest/20digits	143.0 ms	142.7 ms	-0.2%
primetest/30digits	210.3 ms	210.5 ms	+0.1%
primetest/40digits	289.5 ms	291.1 ms	+0.6%
primetest/50digits	276.2 ms	276.4 ms	+0.1%
queens4x4/bt	9.905 ms	9.601 ms	-3.1%
queens4x4/bm	13.94 ms	13.77 ms	-1.2%
queens4x4/bjbt1	12.59 ms	12.33 ms	-2.1%
queens4x4/bjbt2	13.21 ms	12.87 ms	-2.6%
queens4x4/fc	30.99 ms	30.52 ms	-1.5%
queens5x5/bt	128.8 ms	125.7 ms	-2.4%
queens5x5/bm	164.2 ms	161.1 ms	-1.9%
queens5x5/bjbt1	160.0 ms	153.3 ms	-4.2%
queens5x5/bjbt2	163.0 ms	161.2 ms	-1.1%
queens5x5/fc	387.9 ms	382.2 ms	-1.5%

[michaelpj]

/benchmark plutus-benchmark:validation

[github-actions]

Comparing benchmark results of ' plutus-benchmark:validation' on '9fabd96d2' (base) and '6414603b4' (PR)

Results table

Script	9fabd96	6414603	Change
auction_1-1	148.0 μs	151.0 μs	+2.0%
auction_1-2	655.2 μs	650.0 μs	-0.8%
auction_1-3	642.7 μs	640.2 μs	-0.4%
auction_1-4	190.6 μs	195.7 μs	+2.7%
auction_2-1	150.9 μs	153.8 μs	+1.9%
auction_2-2	646.9 μs	650.6 μs	+0.6%
auction_2-3	854.4 μs	862.5 μs	+0.9%
auction_2-4	642.0 μs	641.0 μs	-0.2%
auction_2-5	189.9 μs	196.8 μs	+3.6%
crowdfunding-success-1	178.3 μs	180.8 μs	+1.4%
crowdfunding-success-2	178.0 μs	181.9 μs	+2.2%
crowdfunding-success-3	177.9 μs	181.9 μs	+2.2%
currency-1	235.2 μs	237.4 μs	+0.9%
escrow-redeem_1-1	326.1 μs	330.7 μs	+1.4%
escrow-redeem_1-2	326.6 μs	328.3 μs	+0.5%
escrow-redeem_2-1	379.6 μs	382.7 μs	+0.8%
escrow-redeem_2-2	380.2 μs	386.0 μs	+1.5%
escrow-redeem_2-3	383.7 μs	385.5 μs	+0.5%
escrow-refund-1	132.2 μs	134.7 μs	+1.9%
future-increase-margin-1	236.6 μs	236.6 μs	0.0%
future-increase-margin-2	518.7 μs	519.5 μs	+0.2%
future-increase-margin-3	518.1 μs	519.3 μs	+0.2%
future-increase-margin-4	485.1 μs	488.5 μs	+0.7%
future-increase-margin-5	855.2 μs	856.9 μs	+0.2%
future-pay-out-1	235.3 μs	236.5 μs	+0.5%
future-pay-out-2	517.3 μs	522.6 μs	+1.0%
future-pay-out-3	518.3 μs	517.7 μs	-0.1%
future-pay-out-4	852.9 μs	859.0 μs	+0.7%
future-settle-early-1	235.3 μs	235.9 μs	+0.3%
future-settle-early-2	517.2 μs	518.8 μs	+0.3%
future-settle-early-3	518.0 μs	518.9 μs	+0.2%
future-settle-early-4	634.4 μs	638.4 μs	+0.6%
game-sm-success_1-1	373.9 μs	377.1 μs	+0.9%
game-sm-success_1-2	162.8 μs	167.9 μs	+3.1%
game-sm-success_1-3	635.4 μs	638.0 μs	+0.4%
game-sm-success_1-4	191.6 μs	195.0 μs	+1.8%
game-sm-success_2-1	374.1 μs	376.4 μs	+0.6%
game-sm-success_2-2	163.5 μs	166.9 μs	+2.1%
game-sm-success_2-3	634.6 μs	637.0 μs	+0.4%
game-sm-success_2-4	190.8 μs	195.8 μs	+2.6%
game-sm-success_2-5	639.3 μs	641.4 μs	+0.3%
game-sm-success_2-6	191.0 μs	194.7 μs	+1.9%
multisig-sm-1	388.1 μs	391.5 μs	+0.9%
multisig-sm-2	379.8 μs	380.3 μs	+0.1%
multisig-sm-3	381.9 μs	383.6 μs	+0.4%
multisig-sm-4	380.1 μs	382.5 μs	+0.6%
multisig-sm-5	565.6 μs	566.7 μs	+0.2%
multisig-sm-6	387.1 μs	391.3 μs	+1.1%
multisig-sm-7	377.6 μs	381.2 μs	+1.0%
multisig-sm-8	385.6 μs	386.8 μs	+0.3%
multisig-sm-9	381.9 μs	385.1 μs	+0.8%
multisig-sm-10	565.8 μs	566.4 μs	+0.1%
ping-pong-1	317.1 μs	318.2 μs	+0.3%
ping-pong-2	317.7 μs	319.0 μs	+0.4%
ping-pong_2-1	180.5 μs	181.9 μs	+0.8%
prism-1	137.7 μs	140.3 μs	+1.9%
prism-2	396.6 μs	396.6 μs	0.0%
prism-3	335.0 μs	338.3 μs	+1.0%
pubkey-1	116.6 μs	118.7 μs	+1.8%
stablecoin_1-1	950.9 μs	951.1 μs	+0.0%
stablecoin_1-2	160.5 μs	164.6 μs	+2.6%
stablecoin_1-3	1.089 ms	1.094 ms	+0.5%
stablecoin_1-4	170.4 μs	174.7 μs	+2.5%
stablecoin_1-5	1.375 ms	1.377 ms	+0.1%
stablecoin_1-6	212.3 μs	217.3 μs	+2.4%
stablecoin_2-1	950.5 μs	948.4 μs	-0.2%
stablecoin_2-2	160.6 μs	165.0 μs	+2.7%
stablecoin_2-3	1.087 ms	1.094 ms	+0.6%
stablecoin_2-4	170.4 μs	175.5 μs	+3.0%
token-account-1	172.9 μs	173.5 μs	+0.3%
token-account-2	309.6 μs	311.4 μs	+0.6%
uniswap-1	407.7 μs	408.0 μs	+0.1%
uniswap-2	201.5 μs	201.5 μs	0.0%
uniswap-3	1.771 ms	1.770 ms	-0.1%
uniswap-4	279.6 μs	282.6 μs	+1.1%
uniswap-5	1.138 ms	1.134 ms	-0.4%
uniswap-6	271.5 μs	275.1 μs	+1.3%
vesting-1	334.8 μs	337.0 μs	+0.7%

[L-as]

Rather than doing this as a post-hoc optimisation, why not omit emitting unnecessary delays and forces? That seems like a simpler thing to do.

[L-as]

You could in PIR make one constructor for both type abstractions and lambdas of arbitrary (type) parameters, which would seemingly allow you to emit more optimised UPLC.

[michaelpj]

Rather than doing this as a post-hoc optimisation, why not omit emitting unnecessary delays and forces? That seems like a simpler thing to do.

I disagree with this as a compiler design principle. IME (and I think this is the standard wisdom) it's almost always better to make code-gen as dumb as possible and then optimize away the bad stuff. That makes your code gen simpler (which is important since it's almost always pretty complicated anyway), and your optimizations will benefit you in other places.

[michaelpj]

Pretty modest tbh, slightly disappointing.